Plants & Ecology

Comparing avifauna communities and bird functional diversity of forest and farmland in southwest Ethiopia

by Dries Engelen

Plants & Ecology

Plant Ecology 2012/5 Department of Botany Stockholm University

Comparing avifauna communities and bird functional diversity of forest and farmland in southwest Ethiopia

by Dries Engelen

Supervisor: Kristoffer Hylander

Plants & Ecology

Plant Ecology 2012/5 Department of Botany Stockholm University

Plants & Ecology

Plant Ecology Department of Botany Stockholm University S-106 91 Stockholm Sweden

© Plant Ecology ISSN 1651-9248 Printed by FMV Printcenter Cover: Middle photo: Black-crowned Cranes in the farmland. Other pictures from top left to bottom right: Four forest specialist birds: Zoothera Piaggiae, Mandingoa nitidula, Oriolus monacha & Columba larvata. Photos by Dries Engelen. ABSTRACT - Worldwide degradation and conversion of tropical forests affects many species and their provided ecosystem services. Among them are birds, responsible for pollination, seed dispersal, pest control and scavenging. This study, conducted in southwest Ethiopia, compares species composition and bird functional diversity between forest and homegardens close to and far from forest, both in terms of species numbers and bird abundances. Point counts and mist netting were used to obtain data. While the former method detected more species, abundance data from the latter revealed patterns not observed by just comparing species numbers.

I found that species diversity was lower in forest compared to gardens and that the species composition of both communities was significantly distinct. Whereas forest had more forest specialists, gardens held more forest visitors and species of open country. Close and far gardens did not differ in any aspect, except that abundances of forest generalist birds were somewhat higher close to forest. Regarding bird feeding guilds, I found that granivores and nectarivores were more numerous in gardens, while frugivores were more common in forest. Carnivores and omnivores showed no effect. Insectivore proportions were the same for forest and farmland, but their numbers (including those of all sub-guilds) were higher in gardens.

The Ethiopian forest avifauna is poor in comparison with other Afromontane regions, lacking several, mostly insectivorous genera. This could be the result of an extinction after which its geographic isolation made recolonization difficult, especially for dispersal-limited understory species. Nonetheless, and despite their impoverished state, the results suggest that forest remnants are important for forest-dependent species, being a stronghold for forest specialists and supporting higher numbers of forest generalists in nearby gardens. However, future forest regrowth might be at stake with ongoing agricultural encroachment, because gardens hold fewer frugivores, especially forest specialists, which might affect seed dispersal.

SAMMANFATTNING - Många arter påverkas negativt av avskogning och annan mänsklig påverkan på tropiska skogar, vilket också kan få konsekvenser för de ekosystemtjänster dessa arter levererar. En viktig artgrupp i detta avseende är fåglar, som kan ha betydelse för pollination, fröspridning, naturlig skadedjurskontroll och som asätare. Denna studie, utförd i sydvästra Etiopien, jämför artsammansättning och funktionell diversitet hos fåglar i skogar och trädgårdar både nära och långt ifrån skogen. Både antal arter och antal individer har undersökts. Datamaterialet bygger både på observationer från punktinventering och fåglar fångade med slöjnät. Med den första metoden noterade jag fler arter, medan abundansvärden från den andra metoden gjorde att jag såg mönster som inte syntes vid endast jämförelser av artrikedom.

Jag fann att artrikedomen var lägre i skogen i jämförelse med trädgårdarna, men att artsammansättningen var signifikant skiljd däremellan. Skogarna hade fler skogsspecialister medan trädgårdarna hade fler arter från öppna marker och tillfälliga skogsbesökande arter. Trädgårdar som låg nära eller långt ifrån skogen skiljde sig inte på något sätt utom att individtätheten av skogsgeneralister var något högre i trädgårdar nära skogen. När det gäller uppdelningen av fåglarna utifrån vad de äter så fann jag att fröätande och nektarätande arter var vanligare i trädgårdar medan fruktätande arter var vanligare i skogen. Rovfåglar och allätare uppvisade inget tydligt mönster. Proportionen insektsätande fåglelarter var samma i skogen och trädgårdarna, men det absoluta antalet arter (inklusive alla olika underkategorier) var högre i trädgårdarna.

Den etiopiska skogsfågelfaunan är artfattig i förhållande till andra bergstrakter i Afrika och saknar flera släkten av framförallt insektsätande fåglar. Detta skulle kunna bero på ett tidigare utdöende varefter en senare återkolonisering varit svår på grund av den geografiska isoleringen, speciellt för insektsätande fåglar som huvudsakligen finns i undervegetationen. Trots den relativa artfattigdomen så visar resultatet från min studie hursomhelst att skogarna är viktiga för skogsberoende fåglar. Den största betydelsen har de för skogsspecialister där, men även genom en positiv effekt på abundansen av skogsgeneralister i närliggande trädgårdar. Dock kan framtida återväxt av skogar bli problematisk på grund av en pågående omvandling av skogar till trädgårdar, eftersom trädgårdar hyser färre fruktätande fåglar, vilket kan påverka fröspridning över landskapet.

KEYWORDS: bird diversity, tropical agroecosystem, Ethiopia, feeding guilds, distance to forest, mist netting, point count.

TABLE OF CONTENTS

Abstract & Sammanfattning ………………………………………………………………….. 3

Table of Contents …………………………………………………………………………….. 4

Introduction ………………………………………………………….……………………….. 5

Material & Methods ………………………………………………………………………….. 9

Study area & sites …………………………………………………………………….. 9

Bird surveys & classifications …..…………………………………….…………….. 10

Data analyses ….…………………………………………………………………….. 12

Results ………………………………………………………………………………………. 14

General patterns in the bird community ……...……………………………………... 14

Feeding guilds ………………………………………………………………………. 17

Mist netting vs. point counts ………………………………………………………... 18

Discussion …………………………………………………………………………………... 20

Species composition of forest and farmland ………………………………………... 20

Effect of distance to forest ………………………………………………………….. 22

Feeding guild compositions ………………………………………………………… 23

Comparing bird survey techniques …………………………………………………. 26

Conclusions ………………………………………………………………………………… 28

Acknowledgements ………………………………………………………………………… 29

References ………………………………………………………………………………….. 30

Appendix …………………………………………………………………………………… 34

INTRODUCTION

Habitat degradation and severe ecosystem alterations due to anthropogenic activities is the most important cause for biodiversity losses worldwide (Vitousek et al. 1997). Tropical regions, which harbor the vast majority of this diversity (Gray et al. 2007), are subjected to increasing land-cover changes as a result of accelerating human population growth (Teketay 1992). Worldwide, tropical forests are being logged and degraded because of an increasing demand for forest resources, or are converted into farmland and plantations (Laube et al. 2008). Because degraded and modified habitats make up a growing proportion of the tropics nowadays, it is important to assess their ability to sustain biodiversity (Gray et al. 2007; Laube et al. 2008). Alterations in species richness and composition can also affect the functional diversity of the community (Gray et al. 2007) and changes in provided ecosystem services can, in turn, have an effect on humans again (Clough et al. 2009). One group responsible for a number of ecosystem services is birds, which play an important role in pollination, pest control, seed dispersal and scavenging. A decline in their numbers would therefore mean a decline in their provided services (Şekercioğlu et al. 2004). Of course changes in bird composition may also affect their ecosystem disservices, such as seed-eating.

Synthesizing the results of 57 studies from Asia and the Neotropics on tropical birds specifically, Gray et al. (2007) found that birds from different feeding guilds respond differently to forest disturbance. Whereas granivorous species increase significantly after disturbance, the abundance of frugivores and insectivores significantly decreased. Declines in the numbers of omnivores, carnivores and nectarivores were also observed, though less outspoken because of regional differences. In another study based on global data Tscharntke et al. (2008) found similar results for granivores and insectivores, but instead noticed an increase in (small) frugivores and nectarivores with the conversion of forests to agriculture (until a point when disturbance was so severe that also these groups declined). Furthermore results showed that birds in agricultural habitats are more often generalists.

Overall the increase of granivores and the decline of insectivores and large frugivores with forest modification are most strongly supported (Sodhi et al. 2008). The negative impact on insectivores does, however, differ among the various sub-guilds (Dale et al. 2000) and seems most outspoken for species of the understory and large insectivores in general. Birds of the

5 understory are thought to be so sensitive to disturbance because of their inability to disperse in a non-forest matrix (Newmark 1991; Şekercioğlu et al. 2002).

Whether the same patterns occur in Africa is hard to say since the region is relatively understudied (Gray et al. 2007; Laube et al. 2008; Gove et al. still in press). A study conducted in the Guinea-Congolian forest in southern Cameroon indeed showed similar results for insectivore and granivore richness following the conversion of forest to farmland (Waltert et al. 2005). Nectarivorous birds were also more abundant in agriculture whereas frugivore and omnivore numbers did not differ between habitats. However, in Africa there is a sharp distinction between the lowland forest avifauna and that of the montane regions which occur over 1200 - 1500 m. a.s.l. (Moreau 1966; Romdal & Rahbek 2009). Montane avifaunas differ in trophic structure and taxonomic composition and therefore their response to disturbance may be different (Renjifo 1999).

Nowadays the areas high enough for montane conditions are scattered over Africa and the montane forests even more so, mostly as a result of anthropogenic disturbance (Moreau 1966). Except for the Cameroon highlands and a very small area in Angola, the tropical montane forests of Africa are found on the eastern side of the continent. The Ethiopian Highlands are particularly interesting in this aspect since they comprise over 50% of the African land covered by Afromontane vegetation (Tadesse et al. 2001). Despite a dramatic decline in forest cover during the last decades from 40 (but see McCann 1997) to less than 3 percent (Teketay, 1992), this country still holds some of the larger remnants of tropical Afromontane forests. However, as Ethiopia is one of the poorest countries in the world with a fast-growing population of which 83% still lives in rural areas (Population Reference Bureau 2012), pressure on these last forests is likely to increase strongly (McCann 1995; Cheng et al. 1998).

Research on the effect of encroaching agriculture on the avifauna of these forests is rather scarce. A study performed in the extremely fragmented and rather dry northern part of Ethiopia observed more species and a higher number of unique species in the forest patches than in open fields (Aerts et al. 2008). However it was also noted that true forest specialists were already absent from the region. Another study, conducted in the Ghibe valley, also found most bird species and the highest number of unique species in riparian woodlands when compared to wooded grasslands, oxen-plowed land and tractor-plowed fields and this result was regardless of the season (Wilson et al. 1997). Surprisingly, even though species

6 composition differed among land-use types, no shift was apparent in the distribution of the different feeding guilds. Comparing forest with structurally complex agricultural sites Gove et al. (2008; still in press) found similar results regarding species composition and guild distribution, except for higher numbers of granivores in nectarivores in the farmland. Even when comparing insectivorous sub-guilds no significant difference was observed between the two land types. Numbers of forest specialist birds were lower in farmland, but, contrary to the other two studies, overall species diversity was higher in the agricultural landscape. The authors thus argue that structurally complex agriculture promotes species richness and supports high avian functional diversity (Gove et al. still in press).

Other studies on birds of East African montane forests from Kenya (Borghesio 2008; Laube et al. 2008; Mulwa et al. 2012), Uganda (Şekercioğlu 2002; Naidoo 2004) and Tanzania (Fjeldså 1999) also documented a decrease in forest specialists and an increase in overall species numbers with forest disturbance or conversion. The few studies discussing bird functional diversity do, however, note a decline in (several groups of) insectivores (Fjeldså 1999; Şekercioğlu 2002; Mulwa et al. 2012) and sometimes also frugivores (Borghesio 2008; Kirika et al. 2008). Therefore, it seems odd that similar effects have not been observed in the montane forests of Ethiopia as well (Gove et al. still in press).

In this study I aim to shed more light on bird communities of forest and farmland in Ethiopia by investigating compositional changes for all birds and divided into feeding guilds and habitat preference groups while considering both species richness and bird abundances. Additionally I will compare agricultural sites close and far from forest to see whether distance to forest has an effect on bird functional and species diversity. For forest-dependent birds particularly, proximity to forest has been observed to be important (Laube et al. 2008), even if vegetation complexity and tree density are the primary factors determining bird species composition (Naidoo 2004; Laube et al. 2008). Thus, in order to highlight the distance effect, all farmland sites in this study are of similar complexity.

Furthermore, data is obtained by means of point counts and mist netting and a comparison between both methods is made. The latter has the added advantage of also detecting cryptic, ground-foraging and non-singing birds (Dunn & Ralph 2004) and especially insectivores (Dunn & Ralph 2004), which are of particular interest to this study due to the many different sub-guilds. It is also a useful tool to investigate relative abundances of species (Dunn & Ralph 2004).

As in previous studies conducted in agroecosystems, both in Ethiopia (Wilson et al. 1997; Gove et al. 2008) and in other montane regions of East Africa (Naidoo 2004; Mulwa et al. 2012), I expect to find two distinct avifauna communities in forest and farmland with a decline in the abundance and species richness of forest specialists while species of open country will increase in numbers. These effects are thought to intensify with increasing distance from forest. Furthermore close gardens are hypothesized to be richer in species than sites far from forest, because they are more likely to be visited by foraging forest species (especially in the dry season [Laube et al. 2008]) or by forest canopy species which are temporarily adaptable to more open landscapes (Moreau 1966).

With regard to bird functional diversity I expect the relative abundance of the different feeding guilds to differ between forest and farmland. Especially insectivores and possibly frugivores are thought to decline while granivore species will increase. This shift is expected to strengthen with distance from forest. Among insectivorous sub-guilds the understory foliage gleaners and terrestrial species are assumed to be most sensitive (Fjeldså 1999; Dale et al. 2000; Şekercioğlu et al. 2002; Borghesio 2008).

MATERIAL & METHODS

Study area & sites

The fieldwork was conducted from the end of February to the beginning of May 2012 in the countryside between Agaro and Gera in southwest Ethiopia (07o45’-07o48’N, 36o17’-36o24’E, altitude 1850 - 2100 m). This agricultural landscape is bordered to the north and south by two large forest remnants belonging to the Belete-Gera Forest (Fig. 1).

These moist evergreen forests have an average annual temperature of 20ºC and the annual precipitation ranges from 1,500 to more than 2,000 mm (Friis et al. 2011). Characteristic canopy species include Pouteria adolfi-friederici, Albizia gummifera, Croton macrostachyus, Syzygium guineense, Millettia ferruginea and the nowadays rarer Afrocarpus falcatus (Cheng et al. 1998; Friis et al. 2011). The shrubby understory is diverse as well as the number of epiphytes and lianas. In 1998 the Belete-Gera forest covered an area of 150,000 ha, but its size is continuously being reduced due to agricultural encroachment (Cheng et al. 1998). As most other forest remnants in Ethiopia (Teketay 1992) and Eastern Africa (Borghesio 2008) it is also heavily disturbed by human activities like selective logging, livestock grazing and coffee production (Cheng et al. 1998). Especially the latter was found to be common in the study area and much of the understory of small trees and shrubs frequently cleared to enhance coffee production.

The agricultural landscape is structurally complex and contains shrublands, (wet) grasslands, exotic tree plantations and cultivated areas. The smallholder farmers grow several crops like teff, maize, sorghum, khat, avocado, banana and ensete. Their life fences consist of natural vegetation and planted Euphorbia-species. Scattered across the landscape are remnant forest trees like Ficus, which are used for shading coffee or hanging beehives, but also introduced trees such as several Eucalyptus-species, which mostly serve as firewood or construction material.

Initially the plan was to select nine sites: 3 in forest, 3 gardens close to forest (0 -100 m) and 3 gardens far from forest (1500 – 2000 m). However, due to time constraints it was decided to drop one of the forest sites (Fig. 1). Forest sites were selected to be as undisturbed as possible and at a reasonable distance from the edge. Finally both sites were located 300 - 400 m from the border (as suggested by Dale et al. 2000) and although understory clearing was executed,

9 native shrubs and small trees other than Coffea arabica were still present. Also, because the forest is located on hilly slopes, the selected sites were situated at 100 m higher elevation than those in the gardens.

Agricultural sites were typical smallholder farms with home-garden crops and life fences. They were selected to be structurally complex (e.g. having a variety of trees and bushes) and to be as similar as possible. The selected gardens are part of a larger group of sites used for other projects in this landscape (Lemessa et al. unpublished). All sites were at least 1500 m apart. Distances to forest edge and between sites were calculated with Google Earth version 6.2.2.6613 (2012).

Figure 1: Overview of the research area. Darker color indicates the forest and the lighter the farmland. Colored markings indicate the site locations (Green – Forest; Red – Close homegarden; Orange – Far homegarden). Image from Google Earth 6.2.2.6613 (2012).

Bird surveys & classification

Each site was visited 6 times in total, performing both point counts and mist netting. Two times only netting took place, two times only point counting and two times both methods were executed at the same time. So in total the data of each site consists of 4 netting days and 4 point count days. While point counts took place on moments at least 8 days apart, netting

10 was performed two times two consecutive days (in order not to have to move the equipment every day). Sites were visited during mornings, while afternoons were spend processing the data, mending the nets and shifting the equipment to new sites

For mist netting fourteen 4-shelf nets were used in each site: 12 x 12 m and 2 x 9 m. One of the 9s lost its bottom shelf already on the third day and the rest of the fieldwork was conducted with 3 shelves instead. The nets were opened every day from 6:00 am to 12:00 pm and were only closed during rain, when cattle had to pass or if there was danger for the birds. The total time all nets were closed was 1 day + 4 hours in close gardens, 3.5 hours in far gardens and 0 hours in the forest. The full day nets were closed in one of the close gardens was because the bees of the farmer got agitated and started swarming around the nets, making it impossible to continue working at this site. The remaining two days of mist netting at this site were instead performed in neighboring gardens within 200 m.

Net rounds took place every 30 minutes by two persons working in opposite directions to limit time spend around the nets. After the birds had been extracted they were identified using Redman et al. (2011) and banded with numbered metal rings. Additionally several details were written down like age, sex, wing length, tail length, weight, wing moult, body moult and fat score (in case of a Palearctic migrant). Only once the number of birds caught was so numerous that body and wing moult were not determined to speed up the process. After measuring birds were released close to where they had been caught. Birds that were positively identified but that escaped before ringing were taken up in the data nonetheless. The 119 recaptured individuals were noted down, but not measured again and included only once in further analyses. At all sites bird capture rates were lower on the second day of two consecutives ones as also observed by Dale et al. (2000).

Point counts were done in 4 places in every site. These places were selected in a way that the entire garden was more or less surveyed. Between 7:00 am and 10:30 am each place was visited two times during one morning and every time the point count lasted for 20 minutes in which species, number of individuals and location (tree, bush, ground) were written down. Species observation mostly took place through vision (using an 8x17 binoculars), but also through sound. Overflying species were included in the count. One additional point count was conducted in the close garden where problems had occurred during netting in order to make up for the lost day. For statistical analyses the data of this additional day was included in the observations made on point count day 3 in this site.

Observed birds were grouped into 6 different feeding guilds after consulting ‘The Birds of Africa volumes I-VII’ (Brown et al. 1982; Urban et al. 1986; Fry et al. 1988; Keith et al. 1992; Urban et al. 1997; Fry & Keith 2000; Fry & Keith 2004). These guilds are carnivores, frugivores, granivores, insectivores, nectarivores and omnivores. Whenever a species was equally feeding on two or more food sources it was considered to be an omnivore. If it preferred certain food, it was placed into the corresponding feeding guild. Following Gove et al. (still in press [except the ‘bees’ category]) insectivores were further divided into seven different sub-guilds: aerial species, arboreal foliage gleaners, bark gleaners, pouncers, salliers, terrestrial species and understory foliage gleaners. Definitions of the different insectivorous feeding strategies are described by Gokula & Vijayan (2000).

Birds were also classified according to habitat preference as forest specialist, forest generalist, forest visitor or species of open country. This was done primarily by following Bennun et al. (1996), but also ‘The Birds of Africa’ (Brown et al. 1982; Urban et al. 1986; Fry et al. 1988; Keith et al. 1992; Urban et al. 1997; Fry & Keith 2000; Fry & Keith 2004) and Birdlife International (2012). Species that did not get through the ‘forest-filter’ on www.birdlife.org were considered to be birds of open country. Forest specialists and generalists are sometimes lumped as forest-dependent species.

Data analyses

All statistical tests run on the total dataset have been based on the presence/absence of all species in a certain site. This was the only way in which netting and point count data could be combined. When the results of both bird survey techniques are considered separately, mist netting data can also include the individual abundances of species whereas point count data can incorporate observation frequencies (ranging from 1 to 4 based on the number of days a species was seen). It is explicitly mentioned when calculations are based on data which includes abundances or frequencies. All analyses were executed in R 2.15.1 (R development core team 2012. www.r-project.org) except for the rarefaction analysis which was run using a pre-programmed Excel sheet (Donovan & Welden 2002).

The eight rarefaction curves were based on bird frequency distributions derived from the specific mist netting data for each site. The number of species after 5, 10, 15, …, 40

12 individuals was calculated by taking the mean of 1000 simulations. Curves were rarified down to 40 species because this was the lowest number of birds caught in one of the sites.

The ADONIS test (with 999 permutations) was used to see whether the species composition of each community (forest, close, far) differed. NMDS was run (trymax = 50; permutations = 999) on the same input file to illustrate the ADONIS result with a graph.

In order to compare compositions of feeding guilds and habitat preference groups between communities a χ2-test was used in which the data of each community was based on the lumped data of the representative sites. ‘Farmland’ is the combined data of close and far garden sites. ANOVA was used to analyze differences across and between communities in species richness and bird abundances, and in the proportions and species numbers of feeding guilds and habitat preference groups. A general linear model (GLM) with Poisson distribution was preferred over ANOVA when calculations included count data with low numbers. If the residuals were not normally distributed a Mann-Whitney U-test was run instead. Normality was checked with the Shapiro-Wilk test.

Similarly, a χ2-test was used to examine differences in observed patterns between (net vs. pc) and within (presence/absence vs. abundance, and presence/absence vs. observation frequency) bird survey techniques while an ANOVA compared the detection success (percentage of all observed species detected by one method) of both methods in every community.

RESULTS

General patterns in the bird community

I recorded a total of 107 species from 38 families at the 8 different sites (see Appendix 1). Among these were 12 Palearctic migrants, 2 inter-African migrants and 7 endemic species (six of which have an extending range into Eritrea). Seventeen species belong to the Afrotropical Highlands biome (Birdlife International 2012). With mist nets 1092 birds were caught belonging to 68 different species. During the point counts 98 species were observed. The four vultures are considered to be globally threatened (Birdlife International 2012).

The number of bird species observed in forest sites (37.5 mean ± 1.5 SE) was significantly lower than for close (60.7 ± 1.45) and far gardens (56.3 ± 3.38) (ANOVA p = 0.0039 & p = 0.0096 respectively) while no difference was found between the latter two (p = 0.46). This result was similar when considering the data of mist netting or point counts separately. Comparing rarefied species numbers on the pooled netting data, again shows that species richness in close and far gardens differed significantly from forest sites (ANOVA p = 0.006 & p = 0.014 respectively) but not from each other (p = 0.56) (Fig. 2).

Combining all garden data gave a total of 99 species for farmland, while forest had 44. Among all species there were 11 forest specialist species, 20 forest generalists, 57 forest visitors and 19 species of open country. The bird community in forest sites was significantly different from those in close and far gardens (ADONIS, perm=999, p = 0.017). Close and far gardens did not differ in this aspect (p = 0.62) (Fig. 3). Similar results were again obtained when only looking at mist netting or point count data.

Regarding habitat preference groups, 19% of the avifauna community of farmland consists of open country species and 57% of forest visitors while forest generalists and specialists make up the other 24% (18 and 6 percent respectively). In forest these proportions are significantly different (χ2-test p < 0.001) being 0%, 43%, 32% and 25% respectively. Close and far gardens did not differ in the relative proportion of species habitat preferences (p = 0.75) (Fig. 4a).

Actual numbers of forest specialist species were significantly higher in forest sites (GLM p < 0.001) while forest visitors were more abundant in farmland (p < 0.001). Open country species seem to follow the same trend as the latter (Mann-Whitney U-test p = 0.059). No effect was found for forest generalists (p = 0.86). Close and far gardens did not significantly

20 Close 1 Close 2 15

Close 3 Far 1 10

Far 2 # Species # Far 3 5 Forest 1 Forest 2

0 5 10 15 20 25 30 35 40 # Individuals

Figure 2: Species rarefaction curves (rarefied down to 40 species) for the individual sites based on species abundances obtained through netting.

Figure 3: Ordination plot showing the species composition with respect to the communities (forest, close gardens, far gardens). Green circle: COLA & PSAB. Black circle: CICI, COAR, TALE, EUGL, COCAE, CONA, ZOPI, LASP. Codes listed in Appendix 1.

15 differ in number of species for any of the habitat preference groups (Fig. 4b). However, when looking at the mist netting data and incorporating bird abundances it was shown that significantly more forest generalists were caught in gardens close to forest (mean 65 ± 4.9 SE) compared to far sites (47 ± 2.7) (ANOVA p = 0.033). Birds of the other habitat preference groups did not differ in their abundances between close and far gardens.

Figure 4a) Relative species compositions according to habitat 100% preference for farmland, forest, 90% close and far gardens based on the

80% pooled data for every community. 70% open n = number of species. 60% 50% visitor 40% generalist 30% specialist

Proportion of species of Proportion 20% 10% 0% farmland forest close far (n=99) (n=44) (n=86) (n=81)

40 Figure 4b) Mean number of species per habitat preference 35 group for forest and close and far 30 garden sites. Black lines indicate standard error.

# of species of # 10

0 Specialist Generalist Visitor Open

Forest Close Far

Feeding guilds

Forest and farmland did not differ in feeding guild composition (χ2-test p = 0.45), nor did close and far gardens from one another (p = 0.98). However, when considering the individual proportion of each feeding guild per site it was shown that frugivores form a larger proportion of the bird community in forest sites compared to close and far gardens (ANOVA p < 0.001) while percentages of granivores and nectarivores were significantly higher in gardens (p = 0.0086 and p < 0.001 respectively). For insectivores, omnivores and carnivores no difference was observed (Table 1). Comparing close and far gardens no differences in the percentages of the different feeding guilds were found, except for frugivores which were proportionally more abundant in gardens close to forest (p = 0.020).

The actual numbers of species representing each feeding guild were significantly higher in farmland sites compared to forest for granivores (GLM p = 0.0028), insectivores (GLM p = 0.028) and nectarivores (Mann-Whitney U-test p = 0.034). Omnivores followed a similar trend (Mann-Whitney U-test p = 0.059) while there was no effect on the numbers of carnivores (Mann-Whitney U-test p = 0.59). Frugivores, on the other hand, showed an opposing trend as their numbers seemed higher in forest sites (GLM p = 0.066), which is interesting considering the fact that gardens have more than one and a half times as many bird species (Fig. 5). Regarding the insectivore sub-guilds, all of them were richer in farmland compared to forest (Table 2), but none of them showed a significant response. Also between close and far gardens there were no significant statistical differences.

Table 1: Proportion of each feeding guild for forest, close and far gardens based on the mean of the sites for each community. n = number or sites per community. P-values in the first column are from a 2-way-ANOVA comparing forest and close and far gardens while the p-values in the second are from a one-way-ANOVA that compares close and far gardens. * = based on Mann-Whitney U-test.

Feeding guild Forest p-value Close Far p-value (n = 2) (n = 3) (n = 3) Carnivore 10.7 % 0.080 6.0 % 6.9 % 0.59* Frugivore 13.4 % < 0.001 5.0 % 2.9 % 0.020 Granivore 10.6 % 0.0086 22.5 % 19.0 % 0.14 Insectivore 50.7 % 0.70 47.4 % 49.0 % 0.68 Nectarivore 0 % < 0.001 4.4 % 5.4 % 0.15 Omnivore 14.6 % 0.48 14.8 % 16.7 % 0.36

Table 2: Number of species per insectivore sub-guild for forest, close and far gardens, and farmland based on the pooled data for each community. n = number of sites per community.

Insectivore sub-guilds Forest Close Far Farmland (n = 2) (n = 3) (n = 3) (n = 6) Aerial 3 6 6 6 Arboreal foliage gleaner 7 12 10 14 Bark gleaner 0 3 3 4 Pouncer 0 3 6 6 Sallier 4 9 8 9 Terrestrial 4 5 3 5 Understorey foliage gleaner 4 5 5 6

35 Figure 5: Mean number of species per feeding guild for 30 forest, close and far garden sites. Black lines indicate 25 standard error. C = carnivore; F = frugivore; G = granivore; 20 I = insectivore; N = nectarivore; O = omnivore. 15

# of species of # 10

0 C F G I N O

Forest Close Far

Mist netting vs. point counts

For all sites the number of species observed using point counts was higher than with mist netting, but the latter did add ‘unique’ birds to each site. Point counts recorded between 84 % and 92% of all observed birds in each site and there was no difference among forest, close and far gardens (ANOVA p = 0.16). Mist netting detections, however, were significantly lower in forest sites (36% mean ± 2.8 SE) compared to both close (62% ± 1.7) and far gardens (61% ± 2.1) (ANOVA p < 0.001). This result also came across when looking at the number of individuals that were caught in every site and community: 39 & 46 (= 85) in forest sites; 185, 149 & 169 (= 503) in close gardens; 147, 108 & 249 (= 504) in far gardens.

The feeding guild composition of each community did not differ when using the results of netting or point counts (χ2-test forest p = 0.21; close p = 0.16; far p = 0.69). Also, for both methods there was no difference in the feeding guild patterns (based on number of species) between forest and close (χ2-test net: p = 0.41 & point count: p = 0.20), forest and far (p = 0.83 & p = 0.25) and close and far (p = 0.87 & p = 0.99). However, when incorporating bird abundances in the netting data the feeding guild patterns were significantly different for forest compared to close and to far gardens (χ2-test p < 100% 0.001 & p = 0.0052 respectively), but just not 90% between the 2 gardens types (p = 0.052). The 80% same happened with the point count results once 70% the observation frequencies were taken into O account (p < 0.001 & p < 0.001 & p = 0.35). 60% N 50% Additionally for netting data, when looking at the I 40% full dataset, it was shown that the feeding guild G 30%

composition based on only species presence/ F Proportion of species of Proportion 20% absence data differed significantly from when the C abundances of all these species were included (χ2- 10% test p < 0.001). For instance, granivores made up 0% 25% of the total species caught while their individuals represented 37% of all netted birds.

For omnivores these numbers were 12% and Figure 6: Comparison of feeding guild 21%. On the contrary, only 34% of all captured compositions according to species numbers and to bird abundances (based on combined birds were insectivores, even though 54% of the mist netting data of all sites). C = carnivore; species caught belonged to this feeding guild F = frugivore; G = granivore; I = insecti- (Fig. 6). vore; N = nectarivore; O = omnivore.

DISCUSSION

Species composition of forest and farmland

A notable result of this study is that species richness of forest sites is lower than that of gardens with more than one and a half times as many species observed in the latter. This seems coherent with other studies conducted in agroecosystems of East Africa (Naidoo 2004; Gove et al. 2008; Laube et al. 2008; Mulwa et al. 2012) and again shows the high conservation value of structurally heterogeneous tropical farmland (Tscharntke et al. 2008). The fact that Waltert et al. (2005) did not find such a pattern in West Africa could be due to a lower complexity of the agricultural landscape (Mulwa et al. 2012) or because of a different life-history of that region (Fjeldså 1999). The species pool of non-forest birds in West Africa is generally poorer, probably as a result of limited speciation due to the region’s climatic instability and simple topography (Moreau 1966).

The fact that forest sites in this study harbored low numbers of species could be related to encroaching coffee cultivation (Gove et al. 2008). Furthermore the large difference in species richness in comparison with farmland is also likely to be partly a result of methodological issues such as the larger number of farmland sites or the higher detectability and capture rates of birds in gardens. Nonetheless, similar results have also been obtained in studies with longer sampling periods and a higher number of sites (Mulwa et al. 2012). The observed pattern seems, thus, not just of a methodological nature, but also in need of an ecological explanation. One hypothesis is that East Africa has a rich species pool of birds that are adapted to open landscapes, like savannah, which colonize the forest areas after conversion and disturbance (Mulwa et al. 2012; Gove et al. still in press). Also the fact that highland forests in Ethiopia are poorer in species compared to those in other East African countries (Moreau 1966) might make the difference with the surrounding farmland even more striking. While the Ethiopian highlands were once important stepping stones between Asia, the Middle East and Central Africa it is thought that climatic changes have wiped out most of the countries’ forest avifauna after which recolonization took place from Kenya (Moreau 1966). This was however difficult due to the geographical isolation of the highlands, surrounded by arid and semi-arid lowlands, and has thus resulted in an impoverished avifauna compared to other Afromontane forests (Moreau 1966; Tadesse et al. 2001), even lacking several genera that occur in neighboring countries (Moreau 1966).

While this may explain the general low diversity of birds in forests in Ethiopia, it is still interesting that Gove et al. (2008) observe a substantially larger number of species in forest (72 vs. 44 in this study) while farmland diversity is more or less the same (94 vs. 99). Reasons for this could be the higher number of forest sites in their study or that sampling took place during the wet season. At this time most birds are breeding and may be more restricted to forest (Şekercioğlu 2002). However, the fact that Gove et al. (2008) also observed open country species in their forest sites indicates that their study took place in more degraded forest and/ or closer to the forest edge.

As expected, and a result shared with Gove et al. (2008) and other East African studies (Naidoo 2004; Mulwa et al. 2012), forest and farmland are characterized by two significantly distinct bird communities. This difference was also reflected when birds were grouped after habitat preference. While percentages of forest specialists and generalists were higher in forest, forest visitors and open country species were proportionally more abundant in farmland. Birds assigned to the latter group were even completely absent from forest sites. The sensitivity of especially forest specialist species to forest disturbance and conversion has been documented extensively (Fjeldså 1999; Şekercioğlu 2002; Naidoo 2004; Borghesio 2008; Gove et al. 2008; Kirika et al. 2008; Laube et al. 2008; Mulwa et al. 2012). In this study 5 of the 11 forest specialists were never observed outside the forest; 4 were only observed once (3 of which in a close garden which could be considered to be more part of the forest edge than close to it); and 1 species was observed two times, but in both case only flying over. The only forest specialist that was observed several times in both close and far gardens was the Olive Sunbird. So even if the heterogeneous farmland boasts an enormous number of species it is almost unable to support forest specialists and will thus not offset forest loss (see also Naidoo 2004; Laube et al. 2008).

On the contrary, forest generalists were seen regularly in the agricultural landscape with 18 of the 20 species observed. Interestingly, in the forest itself only 14 of the 20 species were observed. However, it should be borne in mind that simple presence/absence records of forest- dependent species in farmland might not indicate the actual suitability of agricultural habitats to sustain these species in the absence of surrounding forest (Naidoo 2004). Also, outside study sightings of Narina Trogon, Brown Woodland Warbler, Bat Hawk and African Wood Owl in forest sites, indicate that the forest harbors more forest-dependent species than

21 recorded during the study. Again the lower number of forest sites and lower detectability in forest could be responsible for this undersampling. The fact that the study took place during the dry season could have affected this result too, since forest generalists use the forest mostly for breeding (Bennun et al. 1996). Gove et al. (2008) observed birds during the wet season and did indeed find more forest generalists in the forest. However, a personal observation of breeding Montane White-eyes in the farmland might show the potential of the heterogeneous agricultural landscape to maintain (part of) the forest generalists, providing that habitat complexity is not reduced (Aerts et al. 2008).

Effect of distance to forest

Contrary to my expectations distance to the nearest forest had few significant effects. When considering the total dataset close and far gardens did not differ in any aspect, except for a small but significant difference in the proportion of frugivores, and were in most cases even surprisingly similar. Even though vegetation complexity in terms of vertical heterogeneity and tree density has been widely reported as the main factor determining bird diversity and abundance (Wilson et al. 1997; Tadesse et al. 2001; Şekercioğlu 2002; Naidoo 2004; Aerts et al. 2008; Gove et al. 2008; Laube et al. 2008; Mulwa et al. 2012), effects of proximity to the nearest forest have been documented (Naidoo 2004; Gove et al. 2008; Laube et al. 2008). Changes in the community composition and a decline of forest specialist species were the major observations. While this study has similar findings, they do not occur along a distance gradient, but rather immediate at the forest border.

Overall forest specialists were more or less absent from the farmland, suggesting that the vegetation complexity has already decreased to such an extent that distance is not important anymore. For them distance might play a bigger role when it involves moving to secondary forests where tree density and vertical heterogeneity are much higher than in complex gardens (Naidoo 2004). The importance of forest remnants as a species pool for neighboring regenerating forests has been reported more often (Newmark 1991; Naidoo 2004; Aerts et al. 2008; Mulwa et al. 2012).

A reason that no differences were observed in species diversity and composition between close and far gardens could be that the distance of far gardens from forest was not big enough (only 1.5 km). Also the boundary between degraded forest edge and well-wooded agricultural

22 landscape can be difficult to define, which makes it trickier to estimate distances. This might explain why the far garden with the most questionable distance to forest had the highest species richness of the three. It is also possible that the general heterogeneity of the landscape increased connectivity and facilitated species movements.

However, when abundance data from mist netting was included it became obvious that there was one important difference between close and far gardens, namely that forest generalist birds are more plentiful in homegardens close to forest. So while species diversity is similar for close and far gardens, the former holds higher abundances of forest generalists. Forest patches provide high-quality resources that open habitats cannot provide, like nest sites (Aerts et al. 2008). This could be a reason that forest generalists are more common closer to forest. It also illustrates a previously mentioned point very well: simple presence/absence data might overlook the actual suitability of a habitat to harbor certain species. Thus, forest remnants do not only serve as a stronghold for forest specialists, but their presence also supports higher numbers of forest generalists in agricultural areas in their vicinity.

Feeding guild compositions

This study finds no statistically significant differences in the total feeding guild composition of forest and (close and far) gardens, but when considering the guilds separately patterns are observed. Both the actual species numbers and the percentages of granivores and nectarivores were higher in gardens. Increases in these guilds with forest disturbance and conversion have been reported before for Africa (Dale et al. 2000; Waltert et al. 2005; Borghesio 2008), also for Ethiopia (Gove et al. still in press). Higher species numbers of granivores in the farmland are likely related to bigger dominance of wild and cultivated grasses (Waltert et al. 2005). Nectarivores may be more abundant due to the many flowers found in gardens. However, Africa has many forest related nectarivores (Borghesio 2008) and it might be due to their small size and thin vocalizations that they are difficult to detect in the forest canopy (Waltert et al. 2005). This could be a reason that no nectarivores were observed in the forest during this study. Only the nectar-feeding Olive Sunbird was seen, but this has been classified as an omnivore for it also eats insects and berries.

Frugivorous birds, on the other hand, follow an opposite pattern and their percentages are higher in forest compared to farmland. Even the actual species numbers follow this trend,

23 which is interesting considering the fact that the forest harbors significantly fewer species overall. The decline aggravates in far gardens, which on average had an even lower percentage of frugivore species than close sites (although overall species numbers were similar). While other studies find no effects on overall frugivore richness (Waltert et al. 2005; Laube et al. 2008; Gove et al. still in press), some noted a decline of forest specialist frugivores with distance from forest (Laube et al. 2008) and forest disturbance (Borghesio 2008; Kirika et al. 2008). Of the 9 frugivores observed in this study (Table 3) the majority was indeed forest specialist (4) or generalist (3) and this probably explains their higher numbers in forest sites. Additionally, some fruit-eating birds (e.g. Barbets, Tinkerbirds, Thrushes) were classified as omnivores in this study, although none of them is a forest specialist. Frugivore birds play an important role in seed dispersal and subsequent habitat regeneration (Şekercioğlu et al. 2004). Fruit removal by forest specialist frugivores is especially crucial for forests, since a loss in their services is not replaced by other frugivores and fruit-eating birds (Kirika et al. 2008). A decline in overall frugivores, and forest specialist frugivores in particular, with the conversion of forest to farmland could therefore have serious negative impacts on future forest recovery. Although frugivore distribution has also been shown to depend on body mass (Renjifo 1999), season (Mulwa et al. still in press) and (fruiting) tree scatter (Laube et al. 2008) these effects have not been tested in this study. However, ranking the frugivores after body mass shows that the three smallest species were only observed in farmland whereas 4 of the 6 larger species were only seen in forest (Table 3). This perhaps supports the proposed negative impact of forest conversion on mostly larger frugivores (Sodhi et al. 2008).

Table 3: List of frugivores ranked by body mass, showing their occurrence in forest and farmland. Body masses from ‘The Birds of Africa’. * = only overflying

Common Name Habitat Body Mass (g) Forest Farmland Silvery-cheeked Hornbill Generalist 920 Yes Yes African Olive Pigeon Specialist 370-430 Yes - White-cheeked Turaco Specialist 200-315 Yes - African Green Pigeon Generalist 210 Yes Yes Yellow-fronted Parrot Specialist 140-205 Yes Yes* Splendid Starling Specialist 120-150 Yes - Black-winged Lovebird Generalist 53-65 - Yes Speckled Mousebird Visitor 44-62 - Yes Violet-backed Starling Visitor 35-46 - Yes

Carnivores, omnivores and insectivores showed no significant changes across habitats with respect to the relative proportion of each guild per habitat. Although actual species numbers were higher in gardens for all three guilds, the difference with forest was only significant for insectivores, which were twice as numerous in farmland. For insectivores these results seem out of line with many previous studies that highlight the negative impact on insectivorous birds - terrestrial and understory gleaners in particular - caused by forest disturbance (Fjeldså 1999; Dale et al. 2000; Şekercioğlu 2002; Borghesio 2008) and conversion to farmland (Waltert et al. 2005; Mulwa et al. 2012). All sub-guilds had more species in farmland and only 12 of 54 were forest-dependent species. Among these was just one understory foliage gleaner and three terrestrial species, one of which a forest specialist. The other three forest specialists were arboreal foliage gleaners (one mid stratum).

One could think that ongoing forest disturbance (e.g. activities related to coffee cultivation) has already expelled the sensitive understory insectivores from the study sites, but consulting Redman et al. (2011) shows that these species are generally absent among forest specialists in Ethiopia (with the exception of the Abyssinian Ground-Thrush). Understory genera in Kenya such as Illadopsis, Sheppardia and Neocossyphus, but also other forest insectivores that are abundant in other Afromontane forests, like members of the Greenbul family and the genus Apalis, are all lacking in the Ethiopian highland forests. It is possible that many forest species, conceivably especially the dispersal-limited understory insectivores (Şekercioğlu et al. 2002), have not been able to (re)colonize the Ethiopian highlands, traversing miles of arid and semi- arid plains, after the hypothesized extinction of the majority of the forest avifauna (Moreau 1966).

It is important to realize that the obtained results related to bird feeding guilds can be dependent on sampling season. Bird species distributions may change because of breeding requirements or changes in food availability (Mulwa et al. still in press). Especially close to or in the forest edges these changes might be notable (Dale et al. 2000). Another thing is that this study included overflying species in the results. While this method on one side incorporates aerial insectivores, which might be important for the pest control in a certain site, and other species that search for food (such as raptors), it also includes other overflying species which may not have any feeding intentions at all. This could subsequently influence the derived feeding guild composition of a habitat or site.

Finally there is the point that classifying birds into feeding guilds is an arbitrary matter. Even if the diet of most species is quite reasonably documented (for example in ‘the birds of Africa’ [Brown et al. 1982; Urban et al. 1986; Fry et al. 1988; Keith et al. 1992; Urban et al. 1997; Fry & Keith 2000; Fry & Keith 2004]), interpretations may vary. Examples are the Village Weaver, classified as granivore (Waltert et al. 2005) and insectivore (Aerts et al. 2008), and the Copper Sunbird, classified as nectarivore (Waltert et al. 2005) and insectivore (Mulwa et al. 2012). The fact of the matter is that several birds can feed on more than one food item or change their food source depending on the season (e.g. some migratory species) and this makes classifications more variable. However, this may influence interpretation of results strongly. For example, if one site contains 6 obligate frugivores while a second has 6 frugivore-insectivores (classified as frugivores), the interpretation would be that guild functionality does not differ between sites. But is this really the case? Therefore, and for the sake of comparing studies, I think it is important that an even stronger consensus is achieved on bird feeding guild classifications. This will ultimately help to better assess changes in bird functional diversity and their related ecosystem services and disservices.

Comparing bird survey techniques

From the results it became obvious that point count is the preferred method because its success was equal among habitats and it detected over 84% of all observed species in every site. Especially in forest this method was significantly more productive, as also observed in other studies (Whitman et al. 1997; Blake & Loiselle 2001). However, mist netting did add unique species to all of the eight sites and it can thus be argued that a combination of both methods is the best approach when there is enough time and manpower (Whitman et al. 1997), because together they detect the widest range of species (Dunn & Ralph 2004). Whereas point counting might miss nocturnal species, and quiet and secretive species that do not flush easily (Şekercioğlu 2002), mist netting might be biased against ground species, large species, less active species and species that move in levels above the nets (Dale et al. 2000). Especially the latter was an important reason for the lower mist netting success in my forest sites. However, despite their differences in detection success and species bias, feeding guild compositions derived from both methods did not significantly differ, both within and between communities (at least when considering species presence/absence). This result was somewhat surprising for forest sites, since frugivores, an important part of the forest community, were

26 not sampled at all with mist netting. Possibly the overall low number of species in forest gave the statistical analysis too little power.

A drawback of using both methods is that the only way to truly compare their results and to combine them into a bigger dataset is by using the presence/absence data of the observed species. With mist netting and subsequent bird ringing you can make sure you do not count the same individual more than once, while this cannot be done for point count. A problem of using presence/absence data is that a species observed once gets more or less the same value as a species that has been observed more frequently and/ or in larger numbers. When abundances and frequencies are included in the analyses, results may differ from those only based on presence/absence data. This was for instance shown for the feeding guild composition which differed significantly between forest and close and far gardens when calculations were done on abundance or frequency data, but not when only species presence/absence was considered. However, the abundance results may not be representative, because mist netting success was lower in forest sites.

Still, for comparing close and far gardens mist netting results were valid, because both the detection success and the numbers of birds caught were equal for garden types, and proved to be important. Forest generalists seem to be more abundant in close gardens, a result that did not emerge from solely comparing the number of species. Another issue that came to light when including bird abundances was that a feeding guild composition (derived from mist netting data) for all sites combined, differed significantly from one that was based on species presence/absence. The proportion of insectivore species was larger than the percentage of insectivorous individuals while these effects were opposite for granivores and omnivores. Thus, abundance data (obtained through mist netting, while keeping its species bias in mind!) seems important to more accurately determine bird communities in different habitats, affecting both the distribution of habitat preference groups and feeding guild compositions. Therefore the conservation value of certain habitats, as well as shifts in feeding guild related ecosystem services and disservices, might be differently evaluated when bird abundances are taken into account.

CONCLUSIONS

My study demonstrates that this agroecosystem in southwest Ethiopia supports a high number of bird species in which forest and farmland harbor distinct avifaunas. Complex farmland has a higher species diversity than forest and this seems to be the rule rather than the exception for montane regions of Eastern Africa. Yet, in Ethiopia this difference is more striking because of the impoverished state of the forest bird community. The high diversity in farmland again underlines the importance of incorporating complex agroecosystems in conservation policies. However, forest specialists, and to a lesser extent forest generalists, are very dependent on the presence (or proximity) of forests in the landscape. It is thus vital that the last remnants of Afromontane forests are protected from further encroachment and exploitation. More so, because my study also shows that frugivore bird species are among those most sensitive to forest conversion, which could influence seed dispersal services and thus subsequent forest regrowth. While the observed decline of frugivores is in accordance with findings from other tropical regions, a similar sensitivity of insectivore birds has not been observed in this study. Even the various sub-guilds show no difference across the landscape. The Ethiopian highland forests are very poor in species compared to other montane regions, lacking several, mostly insectivorous genera. Possibly this is the result of an extinction after which the geographical isolation of the forests made recolonization difficult, probably especially for dispersal-limited understory species. Further research is needed to investigate this topic. For the other feeding guilds no effects are found for carnivores and omnivores, while granivores and nectarivores are more abundant in farmland. As close and far gardens do not differ in bird functional diversity, it seems that farmers living in complex gardens as far as 1500m from forest do not experience a different subset of ecosystem services and disservices. However, it is very important that greater consensus is reached on the classification of birds in feeding guilds, so that functional diversity can be assessed more precisely and studies be compared more easily. Finally this study shows that point count is the better method to determine species diversity in a site, but that abundance data derived from mist netting can reveal additional patterns, which ultimately helps to better assess the bird community and conservation value of an area.

ACKNOWLEDGEMENTS

This study is part of a larger interdisciplinary project involving both ecologists and human geographers with the aim to study the interaction between farmers and their environment in an agroecosystem in southwest Ethiopia (http://people.su.se/~khyla/mismatch.html). First of all, I want to thank Kristoffer Hylander for giving me an opportunity to participate in this project and for being my supervisor during the thesis work. Furthermore I thank Debissa Lemessa for helping me getting started in the field and for introducing me to all the landowners. I am also happy I could share this experience with all the others of the team that were with me in Agaro.

Special thanks goes out to the people that got up very early every morning and assisted me throughout the fieldwork: driver Belema, field assistants Imam and Kalifa, and Ian Lees who took care of the bird banding. I also like to thank the landowners for allowing me to work on their lands and for their hospitality in general, providing us regularly with a fresh cup of coffee :- ).

Finally I would like to thank family, friends and all other people that helped me or simply listened to my endless talking about birds, in particular Sofia.

The study was supported by a grant (Minor Field Studies) from the Swedish International Development Cooperation Agency (SIDA).

REFERENCES

Aerts, R, Lerouge, F., November, E., Lens, L., Hermy, M. & Muys, B. (2008) Land rehabilitation and the conservation of birds in a degraded Afromontane landscape in northern Ethiopia. – Biodiversity & Conservation 17: 53-69.

Bennun, L., Dranzoa, C. & Pomeroy, D. (1996) The forest birds of Kenya and Uganda. – Journal of East African Natural History 85: 23-48.

Birdlife International (2012) Country profile: Ethiopia. Available from: http://www.birdlife.org/datazone/country/ethiopia. Checked: 2012-08-21.

Blake , J. G. & Loiselle, B. A. (2001) Bird assemblages in second-growth and old-growth forests, Costa Rica: perspectives from mist nets and point counts. – The Auk 118: 304-326.

Borghesio, L. (2008) Effects of human subsistence activities on forest birds in northern Kenya. – Conservation Biology 22: 384-394.

Brown, L. H., Urban, E. K. & Newman, K. (1982) The birds of Africa – Volume I. Academic Press Inc., London.

Cheng, S., Hiwatashi, Y., Imai, H., Naito, M. & Numata T. (1998) Deforestation and degradation of natural resources in Ethiopia: forest management implications from a case study in the Belete-Gera forest. – Journal of Forest Research 3: 199-204.

Clough, Y., Putra, D. D., Pitopang, R. & Tscharntke, T. (2009) Local and landscape factors determine functional bird diversity in Indonesian cacao agroforestry. – Biological Conservation 142: 1032-1041.

Dale, S., Mork, K., Solvang, R. & Plumptre, A. J. (2000) Edge effects on the understory bird community in a logged forest in Uganda. – Conservation Biology 14: 265-276.

Donovan, T. M. & Welden, C. W. (2002) Spreadsheet exercises in conservation biology and landscape ecology. Sinauer Associates, Inc. Publishers, Sunderland, Massachutes. pp. 85-95.

Dunn, E. H. & Ralph, C. J. (2004) Use of mist nets as a tool for bird population monitoring. – Studies in Avian Biology 29: 1-6.

Fjeldså, J (1999) The impact of human forest disturbance on the endemic avifauna of the Udzungwa mountains, Tanzania. – Bird Conservation International 9: 47-62.

Friis, B., Demissew, S. & van Breugel, P. (2011) Atlas of the potential vegetation of Ethiopia. Addis Ababa University Press & Shama Books. pp. 92-105.

Fry, C. H., Keith, S. & Urban, E. K. (1988) The birds of Africa – Volume III. Academic Press Limited, London.

Fry, C. H. & Keith, S (2000) The birds of Africa – Volume VI. Academic Press, London.

Fry, C. H. & Keith, S. (2004) The birds of Africa – Volume VII. Christopher Helm, an imprint of A & C Black Publishers Ltd, London.

Gokula, V. & Vijayan, L. (2000) Foraging pattern of birds during the breeding season in the thorn forest of Mudumalai wildlife sanctuary, Tamil Nadu, southern India. – Tropical Ecology 41: 195-208.

Gove, A. D., Hylander, K., Nemomisa, S. & Shimelis, A. (2008) Ethiopian coffee cultivation – implications for bird conservation and environmental certification. – Conservation Letters 1: 208-216.

Gove, A. D., Hylander, K., Nemomissa, S., Shimelis, A. & Enkossa, W. Structurally complex farms support high avian functional diversity in tropical montane Ethiopia. – Still in press.

Gray, M. A., Baldauf, S. L., Mayhew, P. J. & Hill, J. K. (2007) The response of avian feeding guilds to tropical forest disturbance. – Conservation Biology 21: 133-141.

Keith, S., Urban, E. K. & Fry, C. H. (1992) The birds of Africa – Volume IV. Academic Press Limited, London.

Kirika, J. M., Farwig, N. & Böhning-Gaese, K. (2008) Effects of local disturbance of tropical forests on frugivores and seed removal of a small-seeded Afrotropical tree. – Conservation Biology 22: 318-328.

Laube, I., Breitbach, N. & Böhning-Gaese, K. (2008) Avian diversity in a Kenyan agroecosystem: effects of habitat structure and proximity to forest. – Journal of Ornithology 149: 181-191.

McCann, J. C. (1995) People of the plow – an agricultural history of Ethiopia, 1800-1990. The University of Wisconsin Press. pp.147-190.

McCann, J. C. (1997) The plow and the forest – narratives of deforestation in Ethiopia, 1840- 1992. – Environmental History 2: 138-159.

Moreau, R. E. (1966) The bird faunas of Africa and its islands. Academic Press Inc., London.

Mulwa, R. K., Böhning-Gaese, K. & Matthias Schleuning (2012) High bird species diversity in structurally heterogeneous farmland in western Kenya. – Biotropica 44: 801-809.

Mulwa, R. K. Neuschulz, L., Böhning-Gaese, K. & Matthias Schleuning Seasonal fluctuations of resource abundance and avian feeding guilds across forest-farmland boundaries in tropical Africa. – Still in press.

Naidoo, R. (2004) Species richness and community composition of songbirds in a tropical forest-agricultural landscape. – Animal Conservation 7: 93-105.

Newmark, W. D. (1991) Tropical forest fragmentation and the local extinction of understory birds in the Eastern Usambara Mountains, Tanzania. – Conservation Biology 5: 67-78.

Population Reference Bureau (2012) http://www.prb.org. Checked: 2012-11-09.

Redman, N., Stevenson, T., & Fanshawe, J. (2011) The birds of the Horn of Africa – Second edition. Christopher Helm, an imprint of Bloomsbury Publishing Plc,, London.

Renjifo, L. M. (1999) Composition changes in a Subandean avifauna after long-term forest fragmentation. – Conservation Biology 13: 1124-1139.

Romdal, T. S. & Rahbek, C. (2009) Elevational zonation of Afrotropical forest bird communities along a homogeneous forest gradient. – Journal of Biogeography 36: 327-336.

Şekercioğlu, C. H. (2002) Effects of forestry practices on vegetation structure and bird community of Kibale national park, Uganda. – Biological Conservation 107: 229-240.

Şekercioğlu, C. H., Ehrlich, P. R., Daily, G. C., Aygen, D., Goehring, D. & Sandi, R. F. (2002) Disappearance of insectivorous birds from tropical forest fragments. – Proceedings of the Natural Academy of Sciences 99: 263-267.

Şekercioğlu, C. H., Daily, G. C. & Ehrlich, P. R. (2004) Ecosystem consequences of bird declines. – Proceedings of the Natural Academy of Sciences 101: 18042-18047

Sodhi, N. S., Posa, M. R. C., Lee, T. M. & Warkentin, A. G. (2008) Effect of disturbance or loss of tropical rainforest on birds. – The Auk 125: 511-519.

Tadesse, W. G., Demel, T., Edwards, S. & Olsson, M. (2001) A comparative assessment of diversity and density of birds in Chilimo forest, Ethiopia. – Ostrich Supplement 15: 216-223.

Teketay, D. (1992) Human impact on a natural montane forest in southeastern Ethiopia. – Mountain Research and Development 12: 393-400.

Tscharntke, T., Şekercioğlu, C. H., Dietsch, T. V., Sodhi, N. S., Hoehn, P. & Tylianakis, J. M. (2008) Landscape constraints on functional diversity of birds and insects in tropical agroecosystems. – Ecology 89: 944-951.

Urban, E. K., Fry, C. H. & Keith, S. (1986) The birds of Africa – Volume II. Academic Press Inc., London.

Urban, E. K., Fry, C. H. & Keith, S. (1997) The birds of Africa – Volume V. Academic Press Limited, London.

Vitousek, P.M., Mooney, H.A., Lubchenco, J. & Melillo, J.M. (1997) Human domination of earth’s ecosystems. – Science 277: 494-499.

Waltert, M., Bobo, S., Sainge, M., Fermon, H. & Mühlenberg, M. (2005) From forest to farmland: habitat effects on Afrotropical forest bird diversity. – Ecological Applications 15: 1351-1366.

Whitman, A. A., Hagan, J. M. & Brokaw, N. V. L. (1997) A comparison of two bird survey techniques used in a subtropical forest. – The Condor 99: 955-965.

Wilson, C. J., Reid, R. S., Stanton, N. L. & Perry, B. D. (1997) Effects of land-use and tsetse fly control on bird species richness in southwestern Ethiopia. – Conservation Biology 11: 435- 447.